key: cord-0695877-2ensoxl9
authors: Ling, Z. H.; Guo, H.; Zheng, J. Y.; Louie, P.K.K.; Cheng, H. R.; Jiang, F.; Cheung, K.; Wong, L. C.; Feng, X. Q.
title: Establishing a conceptual model for photochemical ozone pollution in subtropical Hong Kong
date: 2013-09-30
journal: Atmospheric Environment
DOI: 10.1016/j.atmosenv.2012.09.051
sha: ba697a6f915ae0025e33ebbce61efcf35796ea6b
doc_id: 695877
cord_uid: 2ensoxl9

Abstract Photochemical ozone (O3) formation is related to its precursors and meteorological conditions. A conceptual model of O3 air pollution is developed based on the analysis of data obtained at Tung Chung (TC) in Hong Kong. By comparing meteorological parameters between O3 and non-O3 episode days, it was found that high temperatures, strong solar radiation, low wind speeds and relative humidity, northeasterly and/or northwesterly prevailing winds were favorable for the O3 formation, while tropical cyclones were most conducive to the occurrence of O3 episodes. Backward trajectories simulation and graphical illustration of O3 pollution suggested that super-regional (i.e. central and eastern China) and regional (i.e. Pearl River Delta, southern China) transport was another factor that contributed to high O3 levels in Hong Kong. The photochemical O3 formation, generally VOC-limited in Hong Kong, was controlled by a small number of volatile organic compounds (VOCs). Furthermore, the positive matrix factorization (PMF) simulation suggested that solvent usage and vehicular emissions are the major contributors to ambient VOCs in Hong Kong. Finally, this paper presents recommendations for further O3 research and implementation of O3 control strategies.

Ozone (O 3 ), a major component of photochemical smog which impairs visibility and human health, is formed by a complex series of chemical reactions involving volatile organic compounds (VOCs) and nitrogen oxides (NO x ) in the presence of sunlight (Seinfeld and Pandis, 2006; Zheng et al., 2010) . In most urban areas, ambient concentrations of photochemically formed O 3 are related to its precursors, while favorable meteorological conditions are required for the occurrence of high O 3 concentrations (Ding et al., 2004; Guo et al., 2009) . In Hong Kong, high O 3 concentrations or "O 3 episodes" are commonly observed in late summer and autumn, and are closely associated with local photochemical production and longrange transport (Guo et al., 2009; Wang et al., 2009) .

In order to understand the factors that influence photochemical O 3 formation, conceptual models of O 3 air pollution have been developed in recent years for different regions. A conceptual model is a qualitative explanation of the formation and accumulation of O 3 in a given area based on the chemical characteristics of the ambient atmosphere, as well as the physical transport and removal process observed in given locations (Tom et al., 2006; Pun et al., 1998) . Pun et al. (1998) developed a conceptual model to investigate the O 3 formation in San Joaquin Valley in the USA and found that the high O 3 concentrations observed resulted from both the transport of O 3 and precursors from upwind locations, and the local production of O 3 in urban areas within the valley. Tom et al. (2006) developed a conceptual description of the nature of the O 3 air quality problem in the O 3 transport region (OTR), including Connecticut, Delaware, the District of Columbia, Maine, Maryland, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont and north Virginia in USA and concluded that a severe O 3 episode in the OTR can contain elements of long-range air pollution transport from outside the OTR, regional scale transport within the OTR from channeled flows in nocturnal low level jets, and local transport along coastal shores due to bay, lake, and sea breezes.

To formulate and implement effective control strategies for O 3 pollution, the major objective of this study is to develop a conceptual model for the formation, transport and accumulation of O 3 in subtropical Hong Kong by integrated data analysis at Tung Chung (TC) between 2005 and 2010. We chose the TC site because only at this site the most comprehensive dataset including real-time O 3 , CO, NO x , SO 2 , VOCs and meteorological parameters has been systematically collected so far. In addition to the influence of local emission sources, the sampling site is also affected by polluted continental air masses from the highly industrialized PRD region of mainland China (Guo et al., 2009; Zhang et al., 2007) . Thus, this site is capable of monitoring air pollutants transported from the inland PRD region and is suitable for assessing their impact on local air quality. A variety of aspects, including meteorological conditions, source apportionments of O 3 precursors, O 3 -precursor relationships, and the characteristics of air masses in Hong Kong are evaluated. The conceptual model in this study tries to answer the following questions:

(1) what meteorological conditions are favorable to photochemical O 3 formation? (2) Does regional transport have an important influence on high O 3 levels? (3) Is the O 3 formation limited by VOCs, or NO x , or both, and therefore which sources of precursors are the most important ones to be controlled? (4) What are the main sources of O 3 precursors i.e. VOCs and NO x ?

Tung Chung (TC, 22.30 N, 113 .93 E, Fig. 1 ) is a newly-developed residential town located on northern Lantau Island, about 3 km south of the Hong Kong International Airport at Chek Lap Kok with Hong Kong urban center 20 km to the southwest and Macau 38 km to the northeast. The TC site is adjacent to highway and railway lines that connect the airport with other islands of Hong Kong. The potential impact of the airport, highway and railway lines on VOC and NO x levels at the sampling site was demonstrated to be insignificant (AOAQS, 2011; Guo et al., 2007; So and Wang, 2004) . In addition to the influence of local emission sources, TC is also affected by polluted continental air masses from the highly industrialized Pearl River Delta (PRD) region, south China. As such, the TC site is an ideal location to assess the O 3 pollution in Hong Kong.

In this study, real-time VOC data collected from 2005 to 2010 was provided by the Hong Kong Environmental Protection Department (HKEPD). The on-line VOC analyzer (Syntech Spectras GC 955, Series 600/800, Netherland) includes two sampling systems and two column separating systems: GC1 for the 16 C 2 eC 5 hydrocarbons, and GC2 for the 14 C 6 eC 10 hydrocarbons. Details can be found in HKEPD (2012) . In this study, the GC system was operated continuously, with samples collected and analyzed every 30 min. Furthermore, in order to maintain the consistency for the input of different models, the 30-min data was averaged into hourly values.

In addition, built-in computerized programmes of quality control systems, i.e. auto-linearization and auto-calibration, and calibration with span gas were used. Before sampling, the analyzers were calibrated weekly by injecting certified calibration gas (NPL span gas, National Physical Laboratory). In addition, the quality of the real-time data was assured by comparison with the canister samples analyzed by the University of California at Irvine (UCI) (Colman et al., 2001) . The accuracy and precision of the measurements were obtained using the following methods: the accuracy of GC was based on weekly span checks, monthly calibration and annual auto-linearization using NPL gas. The precision of GC was based on quarterly precision check results (the 95% probability limits for the integrated precision based on the weekly precision check results of the latest 3 months). The detection limits of the above VOCs varied compound by compound, ranging from 0.002 to 0.787 ppbv. The accuracy of the measurements was 1e10% for the above VOCs, whereas the precision was 2.5e20%.

Furthermore, other trace gases, including O 3 , CO, SO 2 , NOeNO 2 e NO x and meteorological parameters were also obtained from HKEPD (http://epic.epd.goc.hk/ca/uid/airdata).

2.3. Procedures for developing a conceptual ozone model Fig. 2 illustrates three steps for developing a conceptual ozone model. The first step is to generalize the meteorological conditions, air mass transport characteristics, and precursor levels on O 3 episode days by analyzing the measurement data: first, we identified the O 3 episodes, especially multi-day O 3 episodes from 2005 to 2010; second, we analyzed and compare the meteorological conditions on O 3 and non-O 3 episode days; third, we investigated the characteristics of air masses on O 3 and non-O 3 episode days; fourth, we analyzed the source contributions of VOCs and investigated the O 3 -precursor relationships. Based on the generalized requirements for an O 3 episode day in step 1, we proposed a conceptual model for O 3 pollution which considers atmospheric chemical and physical processes, emission sources of O 3 precursors and meteorological parameters during O 3 episode events (step 2). Once the conceptual model is established, O 3 episode events will be forecast by looking into the meteorological conditions, air mass transport and abundance and sources of precursors (step 3). The proposed conceptual model will be evaluated by case studies.

Hybrid Single-Particle Lagrangian Integrated Trajectory (HYS-LPIT 4.9, http://ready.arl.noaa.gov/HYSPLIT.php), developed by National Oceanic and Atmospheric Administration (NOAA) Air Resource Laboratory, was applied to air masses at TC at every 3 h with the start time of 0000 LT (local time) at the ending point of 200 m above sea level. Furthermore, the trajectories were classified into different groups by using Hierarchical Clustering Method (Ward, 1963) . Details can be found in Guo et al. (2009) .

PMF model v 3.0 was applied to investigate the VOC source apportionments in this study. PMF is a multivariate factor analysis tool that decomposes a matrix of speciated data into two matrices, factor contributions and factor profiles, which can be interpreted by an analyst as to what sources are represented based on observations at the receptor site. A detailed description of this model can be found in Guo et al. (2011) .

The OBM uses concentrations of O 3 and its precursors (i.e. VOCs, CO and NO), as well as meteorological data measured as a function of time at given sites. Detailed description of the OBM model can be found in Ling et al. (2011) . Briefly, the relative incremental reactivity (RIR) functions simulated by the OBM model can be used to evaluate the sensitivity of O 3 photochemical production to the changes in the concentration of its individual precursors in the given areas.

2.4.4. Photochemical trajectory model incorporated with master chemical mechanism (PTM-MCM model)

The PTM is a ground-level Lagrangian box model, simulating complex chemical reactions within a well mixed boundary layer air parcel, which extends from the Earth's surface up to the top of a diurnally varying boundary layer. The chemical mechanism employed in the PTM is an extended version of MCM v3.1, which is a near-explicit chemical mechanism describing the detailed degradation of a large number of emitted organic compounds and the resulting generation of O 3 and other secondary pollutants under conditions appropriate to the atmospheric boundary layer. Furthermore, a photochemical O 3 creation potential (POCP) index is used to describe the relative contribution of VOCs to O 3 formation at the regional scale in the PTM model. The detailed model description and the initial concentrations for the majority of VOCs in the model can be found in Cheng et al. (2010a) . some primary pollutants, i.e. SO 2 and CO, were observed on the O 3 episode days. However, NO level was comparable on both O 3 episode and non-O 3 episode days (p > 0.05), while higher NO 2 (p < 0.01) was observed on O 3 episode days. In addition, temperature and solar radiation were higher on the O 3 episode days than non-O 3 episode days (p < 0.01), while the relative humidity and wind speed were lower (p < 0.01), indicating that meteorological parameters had significant impact on O 3 levels. Indeed, inspection of all the 10 multi-day O 3 episodes found that high O 3 levels were closely associated with high temperature (>28 C at daytime), strong solar radiation (>700 W m À2 at daytime), low wind speed (<2 m s À1 at daytime) and relative humidity (<70% at daytime). On a regional scale, a high-pressure system over China may transport polluted continental air masses to Hong Kong, resulting in high O 3 concentrations. Fig. 4 shows the mean sea level pressure and wind field on 1000 hpa for East Asia from 23 October to 1 December 2007, when five O 3 episode days were observed (Guo et al., 2009) . It is clear that there was an intensive high-pressure system over northern China, while Hong Kong was in the front of the high-pressure ridge. Due to the influence of the highpressure system, the prevailing synoptic winds in Hong Kong were from the northeast, which might lead to high O 3 levels. Indeed, on 24À26 October, an O 3 episode event was found at TC. Furthermore, the diurnal patterns on the O 3 episode days observed in 2005À2010 (Fig. 3) showed a clear diurnal shift in wind speed and direction at TC À southeasterly/northeasterly at lower speeds at night and northerly/northwesterly at higher speeds during daytime when the O 3 levels were usually high, confirming that synoptic winds were associated with high O 3 concentrations. Previous studies demonstrated that the prevailing north and northeast winds brought VOC-laden air and O 3 from inland PRD region to Hong Kong (Guo et al., 2009; Wang et al., 2009 ). shows the typical synoptic charts on the nine O 3 episode events. It is remarkable that the nine severe O 3 episode events from 2005 to 2009 were all influenced by tropical cyclones over the East and South China Sea. The tropical cyclones were also found to be most conducive to the occurrence of high O 3 episodes from 1994 to 2003 (Lee et al., 2002; Huang et al., 2005) . When a tropical cyclone was formed and its center was over the East and the South China Sea, it intensified the inflow in the lower atmospheric layer and the outflow in the upper atmosphere, which caused stagnation and subsidence air over Hong Kong, forming an inversion layer. Such an inversion layer is not favorable to the dispersion of air pollutants. Nevertheless, it should be noted that though all tropical cyclones over the East and the South China Sea caused high O 3 levels, it does not mean that all the O 3 episode days in Hong Kong were induced by tropical cyclones. Under stable meteorological conditions which include high temperature, strong solar radiation, and calm winds, high O 3 levels could also be observed. Huang et al. (2005) counted that about 62% of O 3 episodes from 1999 to 2003 resulted from cyclonic weather patterns.

3.2. Does regional transport have an important influence on high O 3 levels?

Analysis of synoptic wind patterns above suggested the influence of regional transport of air pollutants at TC. In order to determine whether the air masses originated from local, regional and super-regional sources, 24-h backward trajectories were developed using the NOAA-HYSPLIT 4.9 model with the Global Data Assimilation System (GDAS) meteorological data for 3-h intervals at the ending point of 200 m above sea level. These air masses were classified into local, regional (from PRD region), oceanic and super-regional air masses according to their source origins (i.e. longitude and latitude). In addition, cluster analysis was applied to segregate the calculated trajectories into a number of groups for each month from 2005 to 2010 using the hierarchical Ward's method with a square Euclidean measure (Ward, 1963) . In total, 55 cluster groups were obtained. Based on their pathways, air masses arriving at TC were classified into four categories for each month from 2005 to 2010. In order to characterize the four types of air masses, two cases are presented here. Taking September 2005 and June 2010 as examples (Fig. 7) , in September 2005, four categories were described: (i) air masses originating from inland China, passing over Guangdong province and finally arrived at TC (track 1); (ii) air masses originating in the Hong Kong area (track 2); (iii) air mass originating in the South China Sea (track 3) with fast movement; and (iv) air masses originating from the eastern China coast, passing over the coast of eastern Guangdong with very slow movement (track 4). Hence, tracks 1 and 4 were identified as super-regional transport; track 2 was identified as local transport while track 3 originated from the South China Sea. In June 2010, four categories were obtained as well: (i) air masses originating from eastern China coast, passing over the coast of eastern Guangdong with very slow movement (track 1); (ii) air masses originating from the PRD region (track 2) with slow movement; (iii) air mass originating from South China Sea (tracks 3 and 4) with fast movement. Track 1 was identified as super-regional transport, while track 2 was classified as regional Table 1 .

Ozone episode and non-ozone episode days in 2005e2010.

Ozone episode Period Non-ozone episode a Period b transport. The trajectory results confirmed that the air masses were of different origins during different periods. Table 3 presents the average values of different pollutants of the four types of air masses from 2005 to 2010. SO 2 , NO and TVOCs showed higher concentrations in the local and regional air masses, and lower concentrations in the oceanic air masses, most likely due to the dilution of cleaner air from the ocean. However, O 3 had the highest concentration in the super-regional air masses, followed by regional, oceanic and local air masses. The relatively higher O 3 levels in super-regional and regional air masses indicated that longrange transport contributed significantly to the increase of background O 3 levels in Hong Kong. Indeed, Wang et al. (2009) found that long-range transport made a significant contribution to the increase in "total ozone" in urban Hong Kong, and about 81% of the O 3 increase in Hong Kong was due to the O 3 increase in the background air. On the other hand, the lower O 3 level (p < 0.01) in the local air masses was due to the titration of NO at TC. To better understand the NO titration for local air masses, the "total ozone" O x (i.e. O 3 þ NO 2 ) was further examined. The mean concentration of O x was 49.5 AE 2.4 ppbv (mean AE 95% confidence interval) for local air masses, while it was 44.5 AE 0.5 ppbv and 54.7 AE 0.5 ppbv for regional and super-regional air masses, respectively. This confirmed that lower O 3 concentration in local air masses was attributed to high emissions of NO in urban Hong Kong (Guo et al., 2009; Wang et al., 2009) . Further inspection found that over the six years, the transport regime at TC was dominated by the air originating from super-regional transport (about 65% to the total air masses), followed by oceanic air (29%), regional transport (5%) and local emissions (1%). Due to the influence of Asian monsoon circulations, most of the oceanic air arrived at TC in summer, bringing in clean marine air, while super-regional and regional transport were often observed in autumn and winter, leading to the movement of precursor-laden air from the Asian continent to Hong Kong. The high frequency of air masses from super-regional and regional transport is another factor that contributes to high O 3 levels in autumn in Hong Kong, confirmed by the highest O 3 mixing ratio in the super-regional air masses (Table 3) . Indeed, a study conducted at TC in OctoberÀDecember 2007 also found that high O 3 levels were attributed to regional and super-regional transport (Guo et al., 2009; Cheng et al., 2010a) . Fig. 8 shows a conceptualization of the influence of different air masses on O 3 levels at TC. The figure was generated based on the following steps. First, the transport history of air masses was investigated and the air masses were classified for the sampling period. In this study, 24-h backward trajectories were carried out using the HYSPLIT model with the GDAS meteorological data. For each day in the sampling period, eight trajectories were generated corresponding to arrival times at TC of 0000, 0300, 0600, 0900, 1200, 1500, 1800, and 2100 LST at the ending point of 200 m above sea level. These air masses were classified into local, regional, super-regional and oceanic air masses according to their original positions (i.e., latitude and longitude). In addition, the air masses for other hours during the day were classified using the following method: if an air mass at 0300 was identified as regional transport, the air masses at 0200 and 0400 were also considered as regional transport; secondly, the dominant surface winds of different air masses on high O 3 days and non-O 3 episode days in summer and autumn were identified; Finally, the relative O 3 concentrations in different air masses (i.e. local, regional, super-regional and oceanic) with different dominant surface winds were determined. It should be noted that due to the consideration of the statistical power of the trajectory results, a high O 3 day is defined as the day with the highest hourly average O 3 mixing ratio exceeding 80 ppbv in the figure. In addition, only O 3 mixing ratios during daytime (0800À1800, local time) were considered due to the fact that O 3 is formed by VOCs and NO x reacting in the presence of sunlight. It is noteworthy that O 3 episode days usually occur in summer and autumn in Hong Kong. Inspection of the figure found that the dominant surface wind was generally from the northwest during high O 3 days, whereas the prevailing winds were generally from the southwest and northeast during summer (MayÀAugust) and autumn non-O 3 episode days (SeptemberÀNovember), respectively. Moreover, the contributions of super-regional, regional, oceanic and local air masses to the average O 3 levels were 31e49%, 20e31%, 18e29% and 0e27%, respectively, during summer non-O 3 episode days, while they were 29e56%, 19e37%, 15e24% and 0e 31%, respectively, on autumn non-O 3 episode days. The relatively low contribution of oceanic air may be attributed to the fact that the south/southeast winds from the South China Sea brought in clean oceanic air with less primary pollutants and thus led to lower O 3 concentrations (Zheng et al., 2010) . On the other hand, during high O 3 days, super-regional, regional, and local air masses contributed 28e100%, 0e61% and 0e42% respectively to the O 3 mixing ratios in summer, while the respective contributions were 33e100%, 0e56% and 0e39% in autumn. However, no contribution of oceanic air masses was found on high O 3 days in summer and autumn. Overall, regional and super-regional air masses made the most significant contributions to the average O 3 mixing ratio, followed by local and oceanic air masses, consistent with previous studies Zheng et al., 2010) . The results further demonstrated that regional and super-regional air pollution had notable influence on the O 3 pollution in Hong Kong.

Due to the complex O 3 chemistry, photochemical O 3 formation could be dominated by either VOCs or NO x or both. As such, it is important to understand the mechanisms of O 3 formation in a location. Moreover, since O 3 is a secondary pollutant, the regional influence should therefore be considered. On a regional scale, Zheng et al. (2010) reported that the O 3 production was controlled by VOCs in urban areas and possibly controlled by NO x in the northern/northeastern rural areas in the PRD region. Locally, Zhang et al. (2007) and Cheng et al. (2010b) found that the O 3 formation throughout Hong Kong was limited by VOCs, especially by reactive aromatics and some carbonyl compounds, and high NO concentrations suppressed O 3 production. In this section, measurement data for a total of 115 days (i.e. 17, 15, 13, 27, 23 and 20 days in years 2005À2010, respectively) with the maximum hourly O 3 mixing ratios above 80 ppbv were input into the OBM model to investigate the O 3 -precursors relationship at TC. Fig. 9 presents the average RIR values for different precursors on those 115 high O 3 days. It is apparent that O 3 production was generally VOC-limited at TC, indicating that reducing VOCs decreased the O 3 formation and reducing NO could increase O 3 levels. The anthropogenic volatile organic compounds (AVOCs) made the most significant contributions to the O 3 formation, except for 2007 and 2009 when biogenic volatile organic compounds (BVOCs) had the highest contributions due to the high isoprene emissions in these two years. Among the top 10 VOC species, the average RIR value of isoprene was the highest, followed by aromatic compounds i.e. toluene and o-xylene, and alkenes i.e. propene and tran-2-butene (Fig. 10) , revealing that the O 3 formation in Hong Kong was controlled by a small number of VOC species.

In the PTM-MCM model, another method was applied to evaluate the relative contribution of each VOC to O 3 formation, in which the importance of each VOC to O 3 production was ranked by POCP-weighted values (Cheng et al., 2010a) . The POCP values were calculated by the PTM-MCM model, while the POCPweighted values were calculated by combining the POCP value and emissions of each VOC (i.e. the emission rate of the VOC in Hong Kong (tonne yr À1 ) multiplies by its POCP value). The 15 most abundant VOC species accounted for 51.7% of the total VOC emission rates in the PRD region (Cheng et al., 2013) . Among these species, isoprene, cis-2-pentene, 1,3,5-trimethylbenzene, acetaldehyde, 1,2,4-trimethylbenzene, 1,2,3-trimethylbenzene and propene had high POCP values, while toluene, benzene, ethene, isoprene had high emission rates, which are accounted for 6.2%, 5.8%, 5.5%, and 5.3% of the total emission rates, respectively.

After taking account into both the POCP and the emission amount of each VOC (POCP-weighted values), isoprene, ethene, toluene, formaldehyde, m-xylene, propene, acetaldehyde, 1,2,4trimethylbenzene, o-xylene, 1-butene and ethylbenzene became the key precursors to photochemical O 3 formation in Hong Kong (Cheng et al., 2013) . However, some highly reactive species, such as cis-2-pentene, 1,3,5-trimethylbenzene and 1,2,3-trimethylbenzene, had relatively lower contributions to the O 3 formation. This may be due to their low emissions, accounting for 0.1%, 0.5% and 0.4% of the total emission rates, respectively. In contrast, benzene and ethyne a L, R, S, O stand for air masses from local, regional, super-regional and oceanic transport. b Data in the bracket means the total number of air masses observed.

accounted for a relatively high percentage of the total VOC emissions (5.8% and 3.3%). Yet they had negligible contribution to the O 3 formation because of their low reactivity. This feature suggests that the contribution of a VOC to the O 3 formation is determined by the combination of its reactivity and emission. Despite some variations, the results of both OBM and PTM-MCM models showed that some reactive VOCs including BVOCs i.e. isoprene, and AVOCs i.e. toluene, o-xylene and propene had the highest contributions to the O 3 formation. Given that it is difficult to control BVOC emissions, the practical strategy to control O 3 pollution is to effectively reduce AVOC emissions. In addition, POCP-weighted values calculated by PTM-MCM model suggest that the optimal strategy should also consider the emission quantity together with reactivity of individual VOCs when it is formulated and implemented.

3.4. Which emission sources are responsible for the volatile organic compounds in the atmosphere of Hong Kong?

Since photochemical O 3 formation at TC was mostly VOC-limited, investigation of the characteristics of VOC source profiles and apportionments is the prerequisite for the formulation and Fig. 8 . Graphical illustration of O 3 pollution in Hong Kong. The gray, green, blue and red areas represent the relative concentration of O 3 from oceanic, regional, super-regional and local emissions, respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) Table 4 illustrates the source apportionment of VOCs at TC using the PMF model. It can be seen that the main VOC sources were solvent use (e.g. paint and varnish, adhesives and sealants), and household products, gasoline and diesel vehicular emissions, gasoline evaporation, liquefied petroleum gas (LPG) usage, biomass burning, biogenic emissions, and the petrochemical industry. Solvent usage made the greatest contribution to ambient VOCs at TC (41 AE 6 %, mean AE 95% confidence interval), followed by gasoline and diesel vehicular emissions (31 AE 8%) and a mixed source of gasoline evaporation and LPG usage (22 AE 3%). The results are in line with previous studies (Table 4 ). For instance, the contributions of vehicular emissions in this study (21e43%) were similar to previous studies (20e48%), except the study conducted by Lau et al. (2010) who reported the vehicular emissions of w70% in Hong Kong. In contrast, the percentage contribution of vehicular emissions in 2008e2010 obtained in this study was half that quantified in 2001e 2003 (Guo et al., 2006 (Guo et al., , 2007 , perhaps suggesting effective VOC control strategies such as utilizing cleaner diesel for buses in Hong Kong (http://www.epd.gov.hk/epd/english/environmentinhk/air/ air_maincontent.html). On the other hand, the contribution of solvent usage in this study (about 30e56%) was consistent with the results found in previous studies (32e45%), and about twice of the study conducted in 2002e2003 (14e24%). The lower contribution of solvent usage to VOC emissions in 2002e2003 may be attributable to lower usage because of much fewer household, commercial and industrial activities caused by the severe acute respiratory syndrome (SARS) events in 2002e2003 in Hong Kong . Furthermore, results of this study and HKEPD emission inventory (EI) indicated that solvent usage in Hong Kong was still the major contributor to ambient VOCs. In this study, gasoline evaporation and LPG usage were sometimes identified as a mixed source. This mixed source contributed 20.4e27%, similar to the results found in 2001e2003 and 2006e2008 (Guo et al., 2006; Lau et al., 2010) , suggesting the emission from LPG usage and gasoline evaporation had less change in recent years. Therefore, these results point out the importance of effectively controlling solvent usage and vehicular emissions in Hong Kong.

In the sections described above, a conceptual description for O 3 pollution in Hong Kong was developed. During O 3 episodes, the most frequent weather systems affecting Hong Kong were tropical cyclones over the East and the South China Sea (Scenario 1), followed by regional high-pressure systems (anticyclones) to the north over mainland China (Scenario 2), and low-pressure system (trough) to the south and east over the South China Sea (Scenario 3) (Fig. 11aec) (Huang et al., 2006; Guo et al., 2009) . In this section, we conducted the model verification for four cases, i.e. 26 October 2007 , 15 November 2008 and 04 June 2010 as examples for the high O 3 episode scenarios, and 29 July 2010 as an example for non-O 3 episode case. Fig. 12 presents the diurnal variations of meteorological parameters for the four cases. It was found that there were some similarities in weather conditions at TC on the days of 26 October 2007, 04 June and 29 July 2010. First, the temperature and solar radiation were relatively high on these three days, which had daily maximum temperature of 30 C with solar radiation of 806 W m À2 , 29 C with solar radiation of 805 W m À2 , and 33 C with solar radiation of 888 W m À2 , respectively. Second, the relative humidity on these three days was comparable (p > 0.05). However, some differences were also found on these days. Firstly, a tropical cyclone was found over the East China Sea on 26 October 2007 (Fig. 11a) . On 04 June 2010, a low-pressure system (trough) to the south and east This study Lau et al., 2010 Lau et al., 2010 Guo et al., 2011 Guo et al., 2004 Guo et al., 2006 Guo et al., 2007 HKEPD, 2012 was over the South China Sea (Fig. 11c) , while an intense lowpressure system was found over Northern China and Hong Kong was in the front of the low-pressure ridge on 29 July 2010 (Fig. 11d) . Secondly, the wind patterns were different on these three days. On 26 October 2007, the prevailing winds were southeasterly and northeasterly at night and northwesterly during daytime hours, while the dominant winds were southerly on 29 July 2010 with high wind speeds. However, the winds were calm (0.5e2 m s À1 ) on 04 June 2010 with northerly and westerly winds during daytime hours and easterly winds at night. Thirdly, the backward trajectories analysis revealed that air masses arriving at TC were caused by super-regional transport on 26 October 2007, On 15 November 2008, there was an intensive high-pressure system over northern China, while Hong Kong was in the front of the high-pressure ridge. Due to the influence of this high-pressure system, more frequent northerly winds with higher speeds (maximum value: 6 m s À1 ) were observed on 15 November 2008. In addition, the temperature and solar radiation (daily maximum value: 730 W m À2 ) on 15 November 2008 were lower (p < 0.05) and the relative humidity was comparable to those on 29 July 2010. This implied that the O 3 levels could be lower on 15 November 2008 than on 29 July 2010. However, the O 3 mixing ratio was actually higher on 15 November 2008 (hourly peak value: 123 ppbv). Further inspection showed that the discrepancy between O 3 levels on 15 November 2008 and 29 July 2010 was also attributed to the influence of different air masses. The higher O 3 levels on 15 November 2008 were caused by the regional and super-regional transport. Backward trajectory analysis demonstrated that regional and super-regional air masses were frequently observed on 15 November 2008, and the OBM modeling simulations further confirmed that 90% of O 3 was caused by regional/super-regional transport on that day.

In summary, the above discussion indicated that tropical cyclone was mostly conducive to the occurrence of high O 3 mixing ratios. In addition, meteorological conditions such as high temperature, intense solar radiation, low relative humidity and wind speed were favorable to photochemical O 3 formation. Furthermore, polluted continental air masses brought by the northerly winds facilitated the O 3 production in Hong Kong. Nevertheless, it should be noted that these conditions were necessary but insufficient for the occurrence of O 3 episodes.

A conceptual model of O 3 pollution in Hong Kong was developed in this study. Results suggested that tropical cyclone was mostly conducive to the occurrence of high O 3 mixing ratios, while conditions such as high temperature, intense solar radiation, low relative humidity and wind speeds and northeasterly and/or northwesterly prevailing winds were favorable to photochemical O 3 formation. In addition, super-regional and regional transport had significant influence on high O 3 levels in Hong Kong. The simulation of OBM and PTM-MCM models revealed that photochemical O 3 formation was generally VOC-limited in Hong Kong, and solvent usage and vehicular emissions made the most significant contributions to ambient VOCs. It should be noted that the relative contributions of local, regional and super-regional air masses to the O 3 formation on O 3 episode days were extracted from the back trajectory analyses, even when the air movement was stagnant on some O 3 episode days, which could lead to uncertainties on the relative contributions.

By considering the meteorological conditions, atmospheric chemistry and physics, and source apportionment of O 3 precursors, a conceptual description of O 3 pollution problem at TC in Hong Kong was presented above. However, some details remain to be thoroughly understood.

(1) A key factor for the occurrence of O 3 episodes in Hong Kong is the influence of tropical cyclones which cause subsidence, stagnation air and inversion layer. However, the mechanisms of such influence are not fully understood.

(2) Photochemical O 3 formation was generally VOC-limited and related to a small number of VOC species in Hong Kong, and solvent usage and vehicular emissions are the two major VOC sources. It appears that an effective control measure on the emissions of solvent usage and vehicles is an optimal strategy for controlling O 3 pollution in Hong Kong. In addition, the optimal strategy should also consider the emission quantity together with reactivity of individual VOCs when it is formulated and implemented. (3) Many VOC sources are located in the PRD region and regional transport from the inland PRD region has significant influence on the amount of air pollutants in Hong Kong (Tang et al., 2007; Guo et al., 2009) . As such, more concurrent field measurements should be conducted in these two closely interactive areas. In addition, air quality control strategies formulated in Hong Kong should consider the emissions from distant sources together with local sources. (4) In Hong Kong, the combination of the coastline and the mountains gives a terrain with many complex physical features. The role of sea-land breezes in air pollution transport has been well studied previously (Ding et al., 2004) . However, there are relatively few studies on mountain-valley breezes in Hong Kong, though it is very important to air pollution transport in Hong Kong. (5) Although remarkable improvements in the VOC emission inventory and VOC measurements have been made in Hong Kong, significant uncertainties still exist in the source profiles and apportionments of VOC data. Hence, further survey on the VOC emission sources and accurate VOC source profile measurements and analyses are essential to better understand the VOC emissions from different sources in Hong Kong and its surrounding areas i.e. inland PRD region, to provide more reliable results on VOC sources and species which contribute the most to photochemical O 3 formation in Hong Kong. (6) Though photochemical O 3 formation is generally VOC-limited in Hong Kong, the influence of other highly reactive chemicals such as HONO, PAN, H 2 O 2 and other reactive oxidants on the O 3 formation should not be ignored. Besides VOCs, the characteristics of other precursors, i.e. NO x and products, i.e. particles, are required in order to better understand the specific atmospheric chemistry. Flux measurements of O 3 and its precursors over different areas, i.e. Hong Kong and inland PRD, under different meteorological conditions, i.e. anticyclones and cyclones, mixing heights and transport processes are also necessary. (7) Since the conceptual model was developed based on the data collected at one site, i.e. TC, it is suggested that this model should be further confirmed and improved by using data from other sites in Hong Kong.

The content of this paper does not necessarily reflect the views and policies of the Hong Kong Special Administrative Region (HKSAR) Government, nor does mention of trade names or commercial products constitute an endorsement or recommendation of their use.

Episode 1 15e21

Episode 1 10e16

Episode 1 28e31

a Days with low ozone concentration (<200 mg m À3 ), used for comparison with ozone episode day (>200 mg m À3 )

b Ozone episode days excluded if applicable

Final Report prepared by Institute for the Environment

Assessing photochemical ozone formation in the Pearl River Delta with a photochemical trajectory model

On the relationship between ozone and its precursors in the Pearl River Delta: application of an observation-based model (OBM)

Photochemical trajectory modeling of ozone concentrations in Hong Kong

Description of the analysis of a wide range of volatile organic compounds in whole air samples collected during PEM-Tropics A and B

Simulation of sea-land breezes associated with a multi-day episode in the Pearl River Delta of China

Which emission sources are responsible for the volatile organic compounds in the atmosphere of Pearl River Delta

Concurrent observations of air pollutants at two sites in the Pearl River Delta and the implication of regional transport

C1eC8 volatile organic compounds in the atmosphere of Hong Kong: overview of atmospheric processing and source apportionment

Regional and local source contributions to ambient NMHCs of a polluted rural/coastal site in Pearl River Delta

Source apportionment of ambient nonmethane hydrocarbons in Hong Kong: application of a principal component analysis/absolute principal component scores (PCA/APCS) receptor model

Numerical simulation and process analysis of typhoon-related ozone episodes in Hong Kong

Integrated processes analysis and systematic meteorological classification of ozone episodes in Hong Kong

Integrated data analysis and characterization of photochemical ozone in Hong Kong

Source apportionment of ambient volatile organic compounds in Hong Kong. The Science of the Total Environment 408

Ozone episodes in urban Hong Kong

Sources of ambient volatile organic compounds and their contributions to photochemical ozone formation at a site in the Pearl River Delta, southern China

A Conceptual Model for Ozone Formation in the San Joaquin Valley

Atmospheric Chemistry and Physics: From Air Pollution to Climate Change

C3eC12 non-methane hydrocarbons in subtropical Hong Kong: spatialetemporal variations, source receptor relationships and photochemical reactivity

Characteristics and diurnal variations of NMHCs at urban, suburban, and rural sites in the Pearl River Delta and a remote site in South China

The Nature of The Ozone Air Quality Problem in The Ozone Transport Region: A Conceptual Description. Prepared for the Ozone Transport Commission

Increasing surface ozone concentrations in the background atmosphere of Southern China

Hierarchical grouping to optimize an objective function

Ozone production and hydrocarbon reactivity in Hong Kong

Ground-level ozone in the Pearl River Delta region: analysis of data from a recently established regional air quality monitoring network

This project is supported by HKEPD, the Postgraduate Studentship (RPHU) of the Hong Kong Polytechnic University, the Research Grants Council of the Hong Kong Special Administrative Region via grants PolyU5179/09E and N_PolyU545/09. This study is partly supported by internal grants of the Hong Kong Polytechnic University (A-PK25 and 1-ZV7A).